Dec 16 New evaluation completed — GPT-4o achieves 74.4% accuracy on SEC enforcement predictions.

Model Evaluation

Compare AI model performance on real SEC enforcement case predictions.

Model Performance Comparison

GPT-4o Best
64.9%
Overall Accuracy
Resolution38.6%
Monetary53.0%
Injunction78.8%
Officer Bar89.2%
500 cases evaluated
Claude Opus 4
46.8%
Overall Accuracy
Resolution38.6%
Monetary23.5%
Injunction79.8%
Officer Bar92.0%
500 cases evaluated
Gemini 2.0
—
Coming Soon

Showing GPT-4o predictions on 500 evaluated cases below.

View evaluation prompt
You are a legal analyst evaluating SEC enforcement cases. Read the complaint and predict: resolution type (settled/litigated), disgorgement amount, civil penalty, prejudgment interest, has injunction (yes/no), has officer/director bar (yes/no). Respond in JSON format with reasoning.
Matter
Agency
Type
Filed
Status
Score